Aggression in Psychotic Illness

AGGRESSION IN PSYCHOTIC ILLNESS Sir This is with the reference to the letter of Dr.Sureshkumar (1999) in response to our article (Kumar et al 1999). The authors are very thankful that Dr Sureshkumar took the pains to evaluate our article critically However, his concern that the conclusions derived from the study may not be valid seems to be an overcautious view. We would like to address areas of concern raised by him. First of all, regarding the use of BPRS. As the target population was that of patients with psychosis, BPRS is among the best scale. Furthermore, as patients of maina were included only if they had psychotic features and not limited to only mood disturbance and hence they would have scored high on BPRS Secondly, as mentioned by Dr. Sureshkumar, equality of variance is a prerequisite for ANOVA and t-test and it is assumed so and never mentioned In our sample also there was no significant difference in Levene's test for variance among the subgroups. Thirdly, we have used parametric methods as we had a small sample size. Parametric methods have a greater power than non parametric tests and thus the results may be said with greater confidence. Also, our conclusion was based on a significant difference in one finding (ANOVA for BPRS) and no difference in other (t-test for restraint) Dr. Sureshkumar has also used t test for life event score in one of his papers (Sureshkumar, 1998) and recent publications in Indian Journal of Psychiatry have used similar statistics (Singh, 1999, Srivastava,1999 to name a few) Nevertheless, we agree with Dr. Sureshkumar that the use of non parametric tests might have been better especially if we had a larger sample size Lastly, the corrections in chi-square distribution are not done when the expected cell values are more than five in at least 80% of the cells and none of the values are less than one (Cochran criteria, Bland, 1995). Furthermore, they are done when the cell value is small and chi-square is just on the significance level (Kothari, 1990) That is, they are done to avoid the commitment of a type 1 error-rejecting a null hypothesis when it should have been accepted. As we have accepted the null hypothesis, the results would not have been any different As it is we had done the Fishers exact test for restrainment and the exact p value was 0.6581. (1999) …